from __future__ import absolute_import

import copy
import logging
import os
import re
import sys
import xml.sax
from io import open
from typing import (Any, AnyStr, Callable,  # pylint: disable=unused-import
                    Dict, Iterable, List, MutableMapping, MutableSequence,
                    Optional, Set, Tuple, TypeVar, Union, cast)

import requests
from cachecontrol.caches import FileCache
from cachecontrol.wrapper import CacheControl
from rdflib.graph import Graph
from rdflib.namespace import OWL, RDF, RDFS
from rdflib.plugins.parsers.notation3 import BadSyntax
from ruamel import yaml
from ruamel.yaml.comments import CommentedMap, CommentedSeq
from six import StringIO, string_types, iteritems
from six.moves import range, urllib
from typing_extensions import Text  # pylint: disable=unused-import

from . import validate
from .sourceline import SourceLine, add_lc_filename, relname, strip_dup_lineno
from .utils import aslist, onWindows

# move to a regular typing import when Python 3.3-3.6 is no longer supported




_logger = logging.getLogger("salad")
ContextType = Dict[Text, Union[Dict, Text, Iterable[Text]]]
DocumentType = TypeVar('DocumentType', CommentedSeq, CommentedMap)
DocumentOrStrType = TypeVar(
    'DocumentOrStrType', CommentedSeq, CommentedMap, Text)

_re_drive = re.compile(r"/([a-zA-Z]):")

def file_uri(path, split_frag=False):  # type: (str, bool) -> str
    if path.startswith("file://"):
        return path
    if split_frag:
        pathsp = path.split("#", 2)
        frag = "#" + urllib.parse.quote(str(pathsp[1])) if len(pathsp) == 2 else ""
        urlpath = urllib.request.pathname2url(str(pathsp[0]))
    else:
        urlpath = urllib.request.pathname2url(path)
        frag = ""
    if urlpath.startswith("//"):
        return "file:%s%s" % (urlpath, frag)
    return "file://%s%s" % (urlpath, frag)

def uri_file_path(url):  # type: (str) -> str
    split = urllib.parse.urlsplit(url)
    if split.scheme == "file":
        return urllib.request.url2pathname(
            str(split.path)) + ("#" + urllib.parse.unquote(str(split.fragment))
                                if bool(split.fragment) else "")
    raise ValueError("Not a file URI: {}".format(url))

class NormDict(CommentedMap):
    """A Dict where all keys are normalized using the provided function."""

    def __init__(self, normalize=Text):  # type: (Callable) -> None
        super(NormDict, self).__init__()
        self.normalize = normalize

    def __getitem__(self, key):  # type: (Any) -> Any
        return super(NormDict, self).__getitem__(self.normalize(key))

    def __setitem__(self, key, value):  # type: (Any, Any) -> Any
        return super(NormDict, self).__setitem__(self.normalize(key), value)

    def __delitem__(self, key):  # type: (Any) -> Any
        return super(NormDict, self).__delitem__(self.normalize(key))

    def __contains__(self, key):  # type: (Any) -> Any
        return super(NormDict, self).__contains__(self.normalize(key))


def merge_properties(a, b):  # type: (List[Any], List[Any]) -> Dict[Any, Any]
    c = {}
    for i in a:
        if i not in b:
            c[i] = a[i]
    for i in b:
        if i not in a:
            c[i] = b[i]
    for i in a:
        if i in b:
            c[i] = aslist(a[i]) + aslist(b[i])  # type: ignore

    return c


def SubLoader(loader):  # type: (Loader) -> Loader
    return Loader(loader.ctx, schemagraph=loader.graph,
                  foreign_properties=loader.foreign_properties, idx=loader.idx,
                  cache=loader.cache, fetcher_constructor=loader.fetcher_constructor,
                  skip_schemas=loader.skip_schemas, url_fields=loader.url_fields)

class Fetcher(object):
    def fetch_text(self, url):    # type: (Text) -> Text
        raise NotImplementedError()

    def check_exists(self, url):  # type: (Text) -> bool
        raise NotImplementedError()

    def urljoin(self, base_url, url):  # type: (Text, Text) -> Text
        raise NotImplementedError()

    schemes = [u"file", u"http", u"https", u"mailto"]

    def supported_schemes(self):  # type: () -> List[Text]
        return self.schemes


class DefaultFetcher(Fetcher):
    def __init__(self,
                 cache,   # type: Dict[Text, Union[Text, bool]]
                 session  # type: Optional[requests.sessions.Session]
                 ):  # type: (...) -> None
        self.cache = cache
        self.session = session

    def fetch_text(self, url):
        # type: (Text) -> Text
        if url in self.cache and self.cache[url] is not True:
            # treat "True" as a placeholder that indicates something exists but
            # not necessarily what its contents is.
            return cast(Text, self.cache[url])

        split = urllib.parse.urlsplit(url)
        scheme, path = split.scheme, split.path

        if scheme in [u'http', u'https'] and self.session is not None:
            try:
                resp = self.session.get(url)
                resp.raise_for_status()
            except Exception as e:
                raise RuntimeError(url, e)
            return resp.text
        if scheme == 'file':
            try:
                # On Windows, url.path will be /drive:/path ; on Unix systems,
                # /path. As we want drive:/path instead of /drive:/path on Windows,
                # remove the leading /.
                if os.path.isabs(path[1:]):  # checking if pathis valid after removing front / or not
                    path = path[1:]
                with open(urllib.request.url2pathname(str(path)), encoding='utf-8') as fp:
                    return fp.read()

            except (OSError, IOError) as err:
                if err.filename == path:
                    raise RuntimeError(Text(err))
                else:
                    raise RuntimeError('Error reading %s: %s' % (url, err))
        raise ValueError('Unsupported scheme in url: %s' % url)

    def check_exists(self, url):  # type: (Text) -> bool
        if url in self.cache:
            return True

        split = urllib.parse.urlsplit(url)
        scheme, path = split.scheme, split.path

        if scheme in [u'http', u'https'] and self.session is not None:
            try:
                resp = self.session.head(url)
                resp.raise_for_status()
            except Exception:
                return False
            self.cache[url] = True
            return True
        if scheme == 'file':
            return os.path.exists(urllib.request.url2pathname(str(path)))
        if scheme == 'mailto':
            return True
        raise ValueError('Unsupported scheme in url: %s' % url)

    def urljoin(self, base_url, url):  # type: (Text, Text) -> Text
        basesplit = urllib.parse.urlsplit(base_url)
        split = urllib.parse.urlsplit(url)
        if (basesplit.scheme and basesplit.scheme != "file" and split.scheme == "file"):
            raise ValueError("Not resolving potential remote exploit %s from base %s" % (url, base_url))

        if sys.platform == 'win32':
            if base_url == url:
                return url
            basesplit = urllib.parse.urlsplit(base_url)
            # note that below might split
            # "C:" with "C" as URI scheme
            split = urllib.parse.urlsplit(url)

            has_drive = split.scheme and len(split.scheme) == 1

            if basesplit.scheme == "file":
                # Special handling of relative file references on Windows
                # as urllib seems to not be quite up to the job

                # netloc MIGHT appear in equivalents of UNC Strings
                # \\server1.example.com\path as
                # file:///server1.example.com/path
                # https://tools.ietf.org/html/rfc8089#appendix-E.3.2
                # (TODO: test this)
                netloc = split.netloc or basesplit.netloc

                # Check if url is a local path like "C:/Users/fred"
                # or actually an absolute URI like http://example.com/fred
                if has_drive:
                    # Assume split.scheme is actually a drive, e.g. "C:"
                    # so we'll recombine into a path
                    path_with_drive = urllib.parse.urlunsplit(
                        (split.scheme, '', split.path, '', ''))
                    # Compose new file:/// URI with path_with_drive
                    # .. carrying over any #fragment (?query just in case..)
                    return urllib.parse.urlunsplit(
                        ("file", netloc, path_with_drive, split.query, split.fragment))
                if (not split.scheme and not netloc and split.path
                        and split.path.startswith("/")):
                    # Relative - but does it have a drive?
                    base_drive = _re_drive.match(basesplit.path)
                    drive = _re_drive.match(split.path)
                    if base_drive and not drive:
                        # Keep drive letter from base_url
                        # https://tools.ietf.org/html/rfc8089#appendix-E.2.1
                        # e.g. urljoin("file:///D:/bar/a.txt", "/foo/b.txt") == file:///D:/foo/b.txt
                        path_with_drive = "/%s:%s" % (base_drive.group(1), split.path)
                        return urllib.parse.urlunsplit(
                            ("file", netloc, path_with_drive, split.query,
                             split.fragment))

                # else: fall-through to resolve as relative URI
            elif has_drive:
                # Base is http://something but url is C:/something - which urllib would wrongly
                # resolve as an absolute path that could later be used to access local files
                raise ValueError(
                    "Not resolving potential remote exploit %s from base %s"
                    % (url, base_url))

        return urllib.parse.urljoin(base_url, url)


class Loader(object):
    def __init__(self,
                 ctx,                       # type: ContextType
                 schemagraph=None,          # type: Graph
                 foreign_properties=None,   # type: Set[Text]
                 idx=None,                  # type: Dict[Text, Union[CommentedMap, CommentedSeq, Text, None]]
                 cache=None,                # type: Dict[Text, Any]
                 session=None,              # type: requests.sessions.Session
                 fetcher_constructor=None,  # type: Callable[[Dict[Text, Union[Text, bool]], requests.sessions.Session], Fetcher]
                 skip_schemas=None,         # type: bool
                 url_fields=None            # type: Set[Text]
                 ):
        # type: (...) -> None

        normalize = lambda url: urllib.parse.urlsplit(url).geturl()
        if idx is not None:
            self.idx = idx
        else:
            self.idx = NormDict(normalize)

        self.ctx = {}       # type: ContextType
        if schemagraph is not None:
            self.graph = schemagraph
        else:
            self.graph = Graph()

        if foreign_properties is not None:
            self.foreign_properties = set(foreign_properties)
        else:
            self.foreign_properties = set()

        if cache is not None:
            self.cache = cache
        else:
            self.cache = {}

        if skip_schemas is not None:
            self.skip_schemas = skip_schemas
        else:
            self.skip_schemas = False

        if session is None:
            if "HOME" in os.environ:
                self.session = CacheControl(
                    requests.Session(),
                    cache=FileCache(os.path.join(os.environ["HOME"], ".cache", "salad")))
            elif "TMP" in os.environ:
                self.session = CacheControl(
                    requests.Session(),
                    cache=FileCache(os.path.join(os.environ["TMP"], ".cache", "salad")))
            else:
                self.session = CacheControl(
                    requests.Session(),
                    cache=FileCache(os.path.join("/tmp", ".cache", "salad")))
        else:
            self.session = session

        if fetcher_constructor is not None:
            self.fetcher_constructor = fetcher_constructor
        else:
            self.fetcher_constructor = DefaultFetcher
        self.fetcher = self.fetcher_constructor(self.cache, self.session)
        self.fetch_text = self.fetcher.fetch_text
        self.check_exists = self.fetcher.check_exists

        if url_fields is None:
            self.url_fields = set()         # type: Set[Text]
        else:
            self.url_fields = set(url_fields)

        self.scoped_ref_fields = {}     # type: Dict[Text, int]
        self.vocab_fields = set()       # type: Set[Text]
        self.identifiers = []           # type: List[Text]
        self.identity_links = set()     # type: Set[Text]
        self.standalone = None          # type: Optional[Set[Text]]
        self.nolinkcheck = set()        # type: Set[Text]
        self.vocab = {}                 # type: Dict[Text, Text]
        self.rvocab = {}                # type: Dict[Text, Text]
        self.idmap = {}                 # type: Dict[Text, Any]
        self.mapPredicate = {}          # type: Dict[Text, Text]
        self.type_dsl_fields = set()    # type: Set[Text]
        self.subscopes = {}             # type: Dict[Text, Text]

        self.add_context(ctx)

    def expand_url(self,
                   url,                 # type: Text
                   base_url,            # type: Text
                   scoped_id=False,     # type: bool
                   vocab_term=False,    # type: bool
                   scoped_ref=None      # type: int
                   ):
        # type: (...) -> Text
        if url in (u"@id", u"@type"):
            return url

        if vocab_term and url in self.vocab:
            return url

        if bool(self.vocab) and u":" in url:
            prefix = url.split(u":")[0]
            if prefix in self.vocab:
                url = self.vocab[prefix] + url[len(prefix) + 1:]
            elif prefix not in self.fetcher.supported_schemes():
                _logger.warning("URI prefix '%s' of '%s' not recognized, are you missing a $namespaces section?", prefix, url)

        split = urllib.parse.urlsplit(url)

        if ((bool(split.scheme) and split.scheme in [
                u'http', u'https', u'file']) or url.startswith(u"$(")
                or url.startswith(u"${")):
            pass
        elif scoped_id and not bool(split.fragment):
            splitbase = urllib.parse.urlsplit(base_url)
            frg = u""
            if bool(splitbase.fragment):
                frg = splitbase.fragment + u"/" + split.path
            else:
                frg = split.path
            pt = splitbase.path if splitbase.path != '' else "/"
            url = urllib.parse.urlunsplit(
                (splitbase.scheme, splitbase.netloc, pt, splitbase.query, frg))
        elif scoped_ref is not None and not split.fragment:
            pass
        else:
            url = self.fetcher.urljoin(base_url, url)

        if vocab_term and url in self.rvocab:
            return self.rvocab[url]
        else:
            return url

    def _add_properties(self, s):  # type: (Text) -> None
        for _, _, rng in self.graph.triples((s, RDFS.range, None)):
            literal = ((Text(rng).startswith(
                u"http://www.w3.org/2001/XMLSchema#")
                and not Text(rng) == u"http://www.w3.org/2001/XMLSchema#anyURI")
                or Text(rng) == u"http://www.w3.org/2000/01/rdf-schema#Literal")
            if not literal:
                self.url_fields.add(Text(s))
        self.foreign_properties.add(Text(s))

    def add_namespaces(self, ns):  # type: (Dict[Text, Text]) -> None
        self.vocab.update(ns)

    def add_schemas(self, ns, base_url):
        # type: (Union[List[Text], Text], Text) -> None
        if self.skip_schemas:
            return
        for sch in aslist(ns):
            try:
                fetchurl = self.fetcher.urljoin(base_url, sch)
                if fetchurl not in self.cache or self.cache[fetchurl] is True:
                    _logger.debug("Getting external schema %s", fetchurl)
                    content = self.fetch_text(fetchurl)
                    self.cache[fetchurl] = Graph()
                    for fmt in ['xml', 'turtle', 'rdfa']:
                        try:
                            self.cache[fetchurl].parse(data=content, format=fmt, publicID=str(fetchurl))
                            self.graph += self.cache[fetchurl]
                            break
                        except xml.sax.SAXParseException:
                            pass
                        except TypeError:
                            pass
                        except BadSyntax:
                            pass
            except Exception as e:
                _logger.warning("Could not load extension schema %s: %s", fetchurl, e)

        for s, _, _ in self.graph.triples((None, RDF.type, RDF.Property)):
            self._add_properties(s)
        for s, _, o in self.graph.triples((None, RDFS.subPropertyOf, None)):
            self._add_properties(s)
            self._add_properties(o)
        for s, _, _ in self.graph.triples((None, RDFS.range, None)):
            self._add_properties(s)
        for s, _, _ in self.graph.triples((None, RDF.type, OWL.ObjectProperty)):
            self._add_properties(s)

        for s, _, _ in self.graph.triples((None, None, None)):
            self.idx[Text(s)] = None

    def add_context(self, newcontext, baseuri=""):
        # type: (ContextType, Text) -> None
        if bool(self.vocab):
            raise validate.ValidationException(
                "Refreshing context that already has stuff in it")

        self.url_fields = set(("$schemas",))
        self.scoped_ref_fields = {}
        self.vocab_fields = set()
        self.identifiers = []
        self.identity_links = set()
        self.standalone = set()
        self.nolinkcheck = set()
        self.idmap = {}
        self.mapPredicate = {}
        self.vocab = {}
        self.rvocab = {}
        self.type_dsl_fields = set()
        self.subscopes = {}

        self.ctx.update(_copy_dict_without_key(newcontext, u"@context"))

        _logger.debug("ctx is %s", self.ctx)

        for key, value in self.ctx.items():
            if value == u"@id":
                self.identifiers.append(key)
                self.identity_links.add(key)
            elif isinstance(value, MutableMapping) and value.get(u"@type") == u"@id":
                self.url_fields.add(key)
                if u"refScope" in value:
                    self.scoped_ref_fields[key] = value[u"refScope"]
                if value.get(u"identity", False):
                    self.identity_links.add(key)
            elif isinstance(value, MutableMapping) and value.get(u"@type") == u"@vocab":
                self.url_fields.add(key)
                self.vocab_fields.add(key)
                if u"refScope" in value:
                    self.scoped_ref_fields[key] = value[u"refScope"]
                if value.get(u"typeDSL"):
                    self.type_dsl_fields.add(key)
            if isinstance(value, MutableMapping) and value.get(u"noLinkCheck"):
                self.nolinkcheck.add(key)

            if isinstance(value, MutableMapping) and value.get(u"mapSubject"):
                self.idmap[key] = value[u"mapSubject"]

            if isinstance(value, MutableMapping) and value.get(u"mapPredicate"):
                self.mapPredicate[key] = value[u"mapPredicate"]

            if isinstance(value, MutableMapping) and u"@id" in value:
                self.vocab[key] = value[u"@id"]
            elif isinstance(value, string_types):
                self.vocab[key] = value

            if isinstance(value, MutableMapping) and value.get(u"subscope"):
                self.subscopes[key] = value[u"subscope"]

        for k, v in self.vocab.items():
            self.rvocab[self.expand_url(v, u"", scoped_id=False)] = k

        self.identifiers.sort()

        _logger.debug("identifiers is %s", self.identifiers)
        _logger.debug("identity_links is %s", self.identity_links)
        _logger.debug("url_fields is %s", self.url_fields)
        _logger.debug("vocab_fields is %s", self.vocab_fields)
        _logger.debug("vocab is %s", self.vocab)

    def resolve_ref(self,
                    ref,              # type: Union[CommentedMap, CommentedSeq, Text]
                    base_url=None,    # type: Text
                    checklinks=True,  # type: bool
                    strict_foreign_properties=False  # type: bool
                    ):
        # type: (...) -> Tuple[Union[CommentedMap, CommentedSeq, Text, None], Dict[Text, Any]]

        lref = ref           # type: Union[CommentedMap, CommentedSeq, Text, None]
        obj = None           # type: Optional[CommentedMap]
        resolved_obj = None  # type: Optional[Union[CommentedMap, CommentedSeq, Text]]
        inc = False
        mixin = None         # type: Optional[MutableMapping[Text, Any]]

        if not base_url:
            base_url = file_uri(os.getcwd()) + "/"

        sl = SourceLine(obj, None, ValueError)
        # If `ref` is a dict, look for special directives.
        if isinstance(lref, CommentedMap):
            obj = lref
            if "$import" in obj:
                sl = SourceLine(obj, "$import", RuntimeError)
                if len(obj) == 1:
                    lref = obj[u"$import"]
                    obj = None
                else:
                    raise sl.makeError(
                        u"'$import' must be the only field in %s"
                        % (Text(obj)))
            elif "$include" in obj:
                sl = SourceLine(obj, "$include", RuntimeError)
                if len(obj) == 1:
                    lref = obj[u"$include"]
                    inc = True
                    obj = None
                else:
                    raise sl.makeError(
                        u"'$include' must be the only field in %s"
                        % (Text(obj)))
            elif "$mixin" in obj:
                sl = SourceLine(obj, "$mixin", RuntimeError)
                lref = obj[u"$mixin"]
                mixin = obj
                obj = None
            else:
                lref = None
                for identifier in self.identifiers:
                    if identifier in obj:
                        lref = obj[identifier]
                        break
                if not lref:
                    raise sl.makeError(
                        u"Object `%s` does not have identifier field in %s"
                        % (relname(obj), self.identifiers))

        if not isinstance(lref, string_types):
            raise ValueError(u"Expected CommentedMap or string, got %s: `%s`"
                    % (type(lref), Text(lref)))

        if isinstance(lref, string_types) and os.sep == "\\":
            # Convert Windows path separator in ref
            lref = lref.replace("\\", "/")

        url = self.expand_url(lref, base_url, scoped_id=(obj is not None))
        # Has this reference been loaded already?
        if url in self.idx and (not mixin):
            return self.idx[url], {}

        sl.raise_type = RuntimeError
        with sl:
            # "$include" directive means load raw text
            if inc:
                return self.fetch_text(url), {}

            doc = None
            if isinstance(obj, MutableMapping):
                for identifier in self.identifiers:
                    obj[identifier] = url
                doc_url = url
            else:
                # Load structured document
                doc_url, frg = urllib.parse.urldefrag(url)
                if doc_url in self.idx and (not mixin):
                    # If the base document is in the index, it was already loaded,
                    # so if we didn't find the reference earlier then it must not
                    # exist.
                    raise validate.ValidationException(
                        u"Reference `#%s` not found in file `%s`."
                        % (frg, doc_url))
                doc = self.fetch(doc_url, inject_ids=(not mixin))

        # Recursively expand urls and resolve directives
        if bool(mixin):
            doc = copy.deepcopy(doc)
            if doc is not None and mixin is not None:
                doc.update(mixin)
                del doc["$mixin"]
            resolved_obj, metadata = self.resolve_all(
                doc, base_url, file_base=doc_url,
                checklinks=checklinks,
                strict_foreign_properties=strict_foreign_properties)
        else:
            if doc:
                resolve_target = doc
            else:
                resolve_target = obj
            resolved_obj, metadata = self.resolve_all(
                resolve_target, doc_url, checklinks=checklinks,
                strict_foreign_properties=strict_foreign_properties)

        # Requested reference should be in the index now, otherwise it's a bad
        # reference
        if not bool(mixin):
            if url in self.idx:
                resolved_obj = self.idx[url]
            else:
                raise RuntimeError(
                    "Reference `%s` is not in the index. Index contains:\n  %s"
                    % (url, "\n  ".join(self.idx)))

        if isinstance(resolved_obj, CommentedMap):
            if u"$graph" in resolved_obj:
                metadata = _copy_dict_without_key(resolved_obj, u"$graph")
                return resolved_obj[u"$graph"], metadata
            else:
                return resolved_obj, metadata
        else:
            return resolved_obj, metadata

    def _resolve_idmap(self,
                       document,    # type: CommentedMap
                       loader       # type: Loader
                       ):
        # type: (...) -> None
        # Convert fields with mapSubject into lists
        # use mapPredicate if the mapped value isn't a dict.
        for idmapField in loader.idmap:
            if (idmapField in document):
                idmapFieldValue = document[idmapField]
                if (isinstance(idmapFieldValue, MutableMapping)
                        and "$import" not in idmapFieldValue
                        and "$include" not in idmapFieldValue):
                    ls = CommentedSeq()
                    for k in sorted(idmapFieldValue.keys()):
                        val = idmapFieldValue[k]
                        v = None  # type: Optional[CommentedMap]
                        if not isinstance(val, CommentedMap):
                            if idmapField in loader.mapPredicate:
                                v = CommentedMap(
                                    ((loader.mapPredicate[idmapField], val),))
                                v.lc.add_kv_line_col(
                                    loader.mapPredicate[idmapField],
                                    document[idmapField].lc.data[k])
                                v.lc.filename = document.lc.filename
                            else:
                                raise validate.ValidationException(
                                    "mapSubject '{}' value '{}' is not a dict "
                                    "and does not have a mapPredicate.".format(
                                        k, v))
                        else:
                            v = val

                        v[loader.idmap[idmapField]] = k
                        v.lc.add_kv_line_col(loader.idmap[idmapField],
                                             document[idmapField].lc.data[k])
                        v.lc.filename = document.lc.filename

                        ls.lc.add_kv_line_col(
                            len(ls), document[idmapField].lc.data[k])

                        ls.lc.filename = document.lc.filename
                        ls.append(v)

                    document[idmapField] = ls

    typeDSLregex = re.compile(Text(r"^([^[?]+)(\[\])?(\?)?$"))

    def _type_dsl(self,
                  t,        # type: Union[Text, Dict, List]
                  lc,
                  filename):
        # type: (...) -> Union[Text, Dict[Text, Text], List[Union[Text, Dict[Text, Text]]]]

        if not isinstance(t, string_types):
            return t

        m = Loader.typeDSLregex.match(t)
        if not m:
            return t
        first = m.group(1)
        second = third = None
        if bool(m.group(2)):
            second = CommentedMap((("type", "array"),
                                   ("items", first)))
            second.lc.add_kv_line_col("type", lc)
            second.lc.add_kv_line_col("items", lc)
            second.lc.filename = filename
        if bool(m.group(3)):
            third = CommentedSeq([u"null", second or first])
            third.lc.add_kv_line_col(0, lc)
            third.lc.add_kv_line_col(1, lc)
            third.lc.filename = filename
        return third or second or first

    def _resolve_type_dsl(self,
                          document,  # type: CommentedMap
                          loader     # type: Loader
                          ):
        # type: (...) -> None
        for d in loader.type_dsl_fields:
            if d in document:
                datum2 = datum = document[d]
                if isinstance(datum, string_types):
                    datum2 = self._type_dsl(datum, document.lc.data[
                                            d], document.lc.filename)
                elif isinstance(datum, CommentedSeq):
                    datum2 = CommentedSeq()
                    for n, t in enumerate(datum):
                        datum2.lc.add_kv_line_col(
                            len(datum2), datum.lc.data[n])
                        datum2.append(self._type_dsl(
                            t, datum.lc.data[n], document.lc.filename))
                if isinstance(datum2, CommentedSeq):
                    datum3 = CommentedSeq()
                    seen = []  # type: List[Text]
                    for i, item in enumerate(datum2):
                        if isinstance(item, CommentedSeq):
                            for j, v in enumerate(item):
                                if v not in seen:
                                    datum3.lc.add_kv_line_col(
                                        len(datum3), item.lc.data[j])
                                    datum3.append(v)
                                    seen.append(v)
                        else:
                            if item not in seen:
                                datum3.lc.add_kv_line_col(
                                    len(datum3), datum2.lc.data[i])
                                datum3.append(item)
                                seen.append(item)
                    document[d] = datum3
                else:
                    document[d] = datum2

    def _resolve_identifier(self, document, loader, base_url):
        # type: (CommentedMap, Loader, Text) -> Text
        # Expand identifier field (usually 'id') to resolve scope
        for identifer in loader.identifiers:
            if identifer in document:
                if isinstance(document[identifer], string_types):
                    document[identifer] = loader.expand_url(
                        document[identifer], base_url, scoped_id=True)
                    if (document[identifer] not in loader.idx
                            or isinstance(
                                loader.idx[document[identifer]], string_types)):
                        loader.idx[document[identifer]] = document
                    base_url = document[identifer]
                else:
                    raise validate.ValidationException(
                        "identifier field '%s' must be a string"
                        % (document[identifer]))
        return base_url

    def _resolve_identity(self, document, loader, base_url):
        # type: (Dict[Text, List[Text]], Loader, Text) -> None
        # Resolve scope for identity fields (fields where the value is the
        # identity of a standalone node, such as enum symbols)
        for identifer in loader.identity_links:
            if identifer in document and isinstance(
                    document[identifer], MutableSequence):
                for n, v in enumerate(document[identifer]):
                    if isinstance(document[identifer][n], string_types):
                        document[identifer][n] = loader.expand_url(
                            document[identifer][n], base_url, scoped_id=True)
                        if document[identifer][n] not in loader.idx:
                            loader.idx[document[identifer][
                                n]] = document[identifer][n]

    def _normalize_fields(self, document, loader):
        # type: (CommentedMap, Loader) -> None
        # Normalize fields which are prefixed or full URIn to vocabulary terms
        for d in list(document.keys()):
            d2 = loader.expand_url(d, u"", scoped_id=False, vocab_term=True)
            if d != d2:
                document[d2] = document[d]
                document.lc.add_kv_line_col(d2, document.lc.data[d])
                del document[d]

    def _resolve_uris(self,
                      document,  # type: Dict[Text, Union[Text, List[Text]]]
                      loader,    # type: Loader
                      base_url   # type: Text
                      ):
        # type: (...) -> None
        # Resolve remaining URLs based on document base
        for d in loader.url_fields:
            if d in document:
                datum = document[d]
                if isinstance(datum, string_types):
                    document[d] = loader.expand_url(
                        datum, base_url, scoped_id=False,
                        vocab_term=(d in loader.vocab_fields),
                        scoped_ref=loader.scoped_ref_fields.get(d))
                elif isinstance(datum, MutableSequence):
                    for i, url in enumerate(datum):
                        if isinstance(url, string_types):
                            datum[i] = loader.expand_url(
                                url, base_url, scoped_id=False,
                                vocab_term=(d in loader.vocab_fields),
                                scoped_ref=loader.scoped_ref_fields.get(d))


    def resolve_all(self,
                    document,           # type: Union[CommentedMap, CommentedSeq]
                    base_url,           # type: Text
                    file_base=None,     # type: Text
                    checklinks=True,    # type: bool
                    strict_foreign_properties=False  # type: bool
                    ):
        # type: (...) -> Tuple[Union[CommentedMap, CommentedSeq, Text, None], Dict[Text, Any]]
        loader = self
        metadata = CommentedMap()  # type: CommentedMap
        if file_base is None:
            file_base = base_url

        if isinstance(document, CommentedMap):
            # Handle $import and $include
            if (u'$import' in document or u'$include' in document):
                return self.resolve_ref(
                    document, base_url=file_base, checklinks=checklinks,
                    strict_foreign_properties=strict_foreign_properties)
            elif u'$mixin' in document:
                return self.resolve_ref(
                    document, base_url=base_url, checklinks=checklinks,
                    strict_foreign_properties=strict_foreign_properties)
        elif isinstance(document, CommentedSeq):
            pass
        elif isinstance(document, (list, dict)):
            raise Exception("Expected CommentedMap or CommentedSeq, got %s: `%s`" % (type(document), document))
        else:
            return (document, metadata)

        newctx = None  # type: Optional[Loader]
        if isinstance(document, CommentedMap):
            # Handle $base, $profile, $namespaces, $schemas and $graph
            if u"$base" in document:
                base_url = document[u"$base"]

            if u"$profile" in document:
                if newctx is None:
                    newctx = SubLoader(self)
                prof = newctx.fetch(document[u"$profile"])
                newctx.add_namespaces(
                    document.get(u"$namespaces", CommentedMap()))
                newctx.add_schemas(document.get(
                    u"$schemas", []), document[u"$profile"])

            if u"$namespaces" in document:
                if newctx is None:
                    newctx = SubLoader(self)
                newctx.add_namespaces(document[u"$namespaces"])

            if u"$schemas" in document:
                if newctx is None:
                    newctx = SubLoader(self)
                newctx.add_schemas(document[u"$schemas"], file_base)

            if newctx is not None:
                loader = newctx

            if u"$graph" in document:
                metadata = _copy_dict_without_key(document, u"$graph")
                document = document[u"$graph"]
                resolved_metadata = loader.resolve_all(
                    metadata, base_url, file_base=file_base,
                    checklinks=False)[0]
                if isinstance(resolved_metadata, MutableMapping):
                    metadata = resolved_metadata
                else:
                    raise validate.ValidationException(
                        "Validation error, metadata must be dict: %s"
                        % (resolved_metadata))

        if isinstance(document, CommentedMap):
            self._normalize_fields(document, loader)
            self._resolve_idmap(document, loader)
            self._resolve_type_dsl(document, loader)
            base_url = self._resolve_identifier(document, loader, base_url)
            self._resolve_identity(document, loader, base_url)
            self._resolve_uris(document, loader, base_url)

            try:
                for key, val in document.items():
                    subscope = ""  # type: Text
                    if key in loader.subscopes:
                        subscope = "/" + loader.subscopes[key]
                    document[key], _ = loader.resolve_all(
                        val, base_url+subscope, file_base=file_base, checklinks=False)
            except validate.ValidationException as v:
                _logger.warning("loader is %s", id(loader), exc_info=True)
                raise validate.ValidationException("(%s) (%s) Validation error in field %s:\n%s" % (
                    id(loader), file_base, key, validate.indent(Text(v))))

        elif isinstance(document, CommentedSeq):
            i = 0
            try:
                while i < len(document):
                    val = document[i]
                    if isinstance(val, CommentedMap) and (u"$import" in val or u"$mixin" in val):
                        l, import_metadata = loader.resolve_ref(
                            val, base_url=file_base, checklinks=False)
                        metadata.setdefault("$import_metadata", {})
                        for identifier in loader.identifiers:
                            if identifier in import_metadata:
                                metadata["$import_metadata"][import_metadata[identifier]] = import_metadata
                        if isinstance(l, CommentedSeq):
                            lc = document.lc.data[i]
                            del document[i]
                            llen = len(l)
                            for j in range(len(document) + llen, i + llen, -1):
                                document.lc.data[
                                    j - 1] = document.lc.data[j - llen]
                            for item in l:
                                document.insert(i, item)
                                document.lc.data[i] = lc
                                i += 1
                        else:
                            document[i] = l
                            i += 1
                    else:
                        document[i], _ = loader.resolve_all(
                            val, base_url, file_base=file_base, checklinks=False)
                        i += 1
            except validate.ValidationException as v:
                _logger.warning("failed", exc_info=True)
                raise validate.ValidationException("(%s) (%s) Validation error in position %i:\n%s" % (
                    id(loader), file_base, i, validate.indent(Text(v))))

            for identifer in loader.identity_links:
                if identifer in metadata:
                    if isinstance(metadata[identifer], string_types):
                        metadata[identifer] = loader.expand_url(
                            metadata[identifer], base_url, scoped_id=True)
                        loader.idx[metadata[identifer]] = document

        if checklinks:
            all_doc_ids={}  # type: Dict[Text, Text]
            loader.validate_links(document, u"", all_doc_ids,
                                  strict_foreign_properties=strict_foreign_properties)

        return document, metadata

    def fetch(self, url, inject_ids=True):  # type: (Text, bool) -> Any
        if url in self.idx:
            return self.idx[url]
        try:
            text = self.fetch_text(url)
            if isinstance(text, bytes):
                textIO = StringIO(text.decode('utf-8'))
            else:
                textIO = StringIO(text)
            textIO.name = url    # type: ignore
            result = yaml.round_trip_load(textIO, preserve_quotes=True)
            add_lc_filename(result, url)
        except yaml.parser.ParserError as e:
            raise validate.ValidationException("Syntax error %s" % (e))
        if (isinstance(result, CommentedMap) and inject_ids
                and bool(self.identifiers)):
            for identifier in self.identifiers:
                if identifier not in result:
                    result[identifier] = url
                self.idx[self.expand_url(result[identifier], url)] = result
        else:
            self.idx[url] = result
        return result


    FieldType = TypeVar('FieldType', Text, CommentedSeq, CommentedMap)

    def validate_scoped(self, field, link, docid):
        # type: (Text, Text, Text) -> Text
        split = urllib.parse.urlsplit(docid)
        sp = split.fragment.split(u"/")
        n = self.scoped_ref_fields[field]
        while n > 0 and len(sp) > 0:
            sp.pop()
            n -= 1
        tried = []
        while True:
            sp.append(link)
            url = urllib.parse.urlunsplit((
                split.scheme, split.netloc, split.path, split.query,
                u"/".join(sp)))
            tried.append(url)
            if url in self.idx:
                return url
            sp.pop()
            if len(sp) == 0:
                break
            sp.pop()
        if onWindows() and link.startswith("file:"):
            link = link.lower()
        raise validate.ValidationException(
            "Field `%s` references unknown identifier `%s`, tried %s" % (field, link, ", ".join(tried)))

    def validate_link(self, field, link, docid, all_doc_ids):
        # type: (Text, FieldType, Text, Dict[Text, Text]) -> FieldType
        if field in self.nolinkcheck:
            return link
        if isinstance(link, string_types):
            if field in self.vocab_fields:
                if (link not in self.vocab and link not in self.idx
                        and link not in self.rvocab):
                    if field in self.scoped_ref_fields:
                        return self.validate_scoped(field, link, docid)
                    elif not self.check_exists(link):
                        raise validate.ValidationException(
                            "Field `%s` contains undefined reference to `%s`" % (field, link))
            elif link not in self.idx and link not in self.rvocab:
                if field in self.scoped_ref_fields:
                    return self.validate_scoped(field, link, docid)
                elif not self.check_exists(link):
                    raise validate.ValidationException(
                        "Field `%s` contains undefined reference to `%s`"
                        % (field, link))
        elif isinstance(link, CommentedSeq):
            errors = []
            for n, i in enumerate(link):
                try:
                    link[n] = self.validate_link(field, i, docid, all_doc_ids)
                except validate.ValidationException as v:
                    errors.append(v)
            if bool(errors):
                raise validate.ValidationException(
                    "\n".join([Text(e) for e in errors]))
        elif isinstance(link, CommentedMap):
            self.validate_links(link, docid, all_doc_ids)
        else:
            raise validate.ValidationException(
                "`%s` field is %s, expected string, list, or a dict."
                % (field, type(link).__name__))
        return link

    def getid(self, d):  # type: (Any) -> Optional[Text]
        if isinstance(d, MutableMapping):
            for i in self.identifiers:
                if i in d:
                    idd = d[i]
                    if isinstance(idd, string_types):
                        return idd
        return None

    def validate_links(self, document, base_url, all_doc_ids, strict_foreign_properties=False):
        # type: (Union[CommentedMap, CommentedSeq, Text, None], Text, Dict[Text, Text], bool) -> None
        docid = self.getid(document)
        if not docid:
            docid = base_url

        errors = []         # type: List[Text]
        iterator = None     # type: Any
        if isinstance(document, MutableSequence):
            iterator = enumerate(document)
        elif isinstance(document, MutableMapping):
            for d in self.url_fields:
                sl = SourceLine(document, d, Text)
                try:
                    if d in document and d not in self.identity_links:
                        document[d] = self.validate_link(d, document[d], docid, all_doc_ids)
                except (validate.ValidationException, ValueError) as v:
                    if d == "$schemas" or (d in self.foreign_properties and not strict_foreign_properties):
                        _logger.warning(strip_dup_lineno(sl.makeError(Text(v))))
                    else:
                        errors.append(sl.makeError(Text(v)))

            try:
                for identifier in self.identifiers:  # validate that each id is defined uniquely
                    if identifier in document:
                        sl = SourceLine(document, identifier, Text)
                        if document[identifier] in all_doc_ids and sl.makeLead() != all_doc_ids[document[identifier]]:
                            # TODO: Validator should local scope only in which duplicated keys are prohibited.
                            # See also https://github.com/common-workflow-language/common-workflow-language/issues/734
                            # In the future, it should raise validate.ValidationException instead of _logger.warn
                            _logger.warning("%s object %s `%s` previously defined" % (all_doc_ids[document[identifier]], identifier, relname(document[identifier]), ))
                        else:
                            all_doc_ids[document[identifier]] = sl.makeLead()
                            break
            except validate.ValidationException as v:
                errors.append(sl.makeError(Text(v)))

            if hasattr(document, "iteritems"):
                iterator = iteritems(document)
            else:
                iterator = list(document.items())
        else:
            return

        for key, val in iterator:
            sl = SourceLine(document, key, Text)
            try:
                self.validate_links(val, docid, all_doc_ids, strict_foreign_properties=strict_foreign_properties)
            except validate.ValidationException as v:
                if key in self.nolinkcheck or (isinstance(key, string_types) and ":" in key):
                    _logger.warning(validate.indent(Text(v)))
                else:
                    docid2 = self.getid(val)
                    if docid2 is not None:
                        errors.append(sl.makeError(
                            "checking object `%s`\n%s" % (relname(docid2), validate.indent(Text(v)))))
                    else:
                        if isinstance(key, string_types):
                            errors.append(sl.makeError("checking field `%s`\n%s" % (
                                key, validate.indent(Text(v)))))
                        else:
                            errors.append(sl.makeError("checking item\n%s" % (
                                validate.indent(Text(v)))))
        if bool(errors):
            if len(errors) > 1:
                raise validate.ValidationException(
                    u"\n".join(errors))
            else:
                raise validate.ValidationException(errors[0])
        return


D = TypeVar('D', CommentedMap, ContextType)

def _copy_dict_without_key(from_dict, filtered_key):
    # type: (D, Any) -> D
    new_dict = CommentedMap(from_dict.items())
    if filtered_key in new_dict:
        del new_dict[filtered_key]
    if isinstance(from_dict, CommentedMap):
        new_dict.lc.data = copy.copy(from_dict.lc.data)
        new_dict.lc.filename = from_dict.lc.filename
    return new_dict